Evaluating Missing Data Imputation Methods in Binary Datasets

This study evaluates the performance of five imputation methods - Mode, SHD, KNN, RF, and MICE - for binary data across MCAR, MAR, and MNAR mechanisms. Through simulation, we compare precision and classification accuracy. KNN and RF are superior under MCAR and MAR, while RF is most robust under MNAR. These findings demonstrate that selecting the appropriate imputation method based on the missingness mechanism is essential for maintaining data integrity and predictive performance in binary datasets.